Google DeepMind on November 11 released AlphaGenome 2, a deep-learning model trained on more than 100 million individual cell profiles across 68 tissues to predict how DNA variants affect gene regulation in specific cell types and states2,3. The model is available for non-commercial research and in private preview for commercial research.
The predecessor model, AlphaGenome, released last year, predicted molecular activity across broad tissues and cell lines simultaneously from a one-million-letter DNA sequence1. A companion release, AlphaGenome Atlas, offered precomputed predictions for all 9 billion possible single-letter changes in the human genome. AlphaGenome 2 is a separate model, not an extension of that atlas.
The core architectural change: AlphaGenome 2 evaluates one condition at a time, conditioning each prediction on a selected tissue, cell type, or cell state — including conditions generated from a researcher's own single-cell experiment. Nearly every cell in the human body carries the same DNA, but different cell types switch different genes on and off, and those patterns shift depending on what a cell is doing. A variant can matter only in a specialized cell or in a cell activated by disease, leaving its signal diluted in a tissue-level average.
Its readouts include gene expression, RNA splicing, and chromatin accessibility, with CpG DNA methylation added to the original model's capabilities.
Google DeepMind said the model could open paths to treatment by giving scientists a more detailed view of how genetic variants operate in individual cell contexts. The same genetic variant can have very different effects depending on the type and state of the cell, making cell-level resolution a prerequisite for connecting variants to disease mechanisms.
ANALYSIS By conditioning predictions on researcher-supplied single-cell data, AlphaGenome 2 moves beyond a fixed catalog of tissues toward a tool that can adapt to the specific cellular contexts a lab is studying — a design that could make the model more immediately useful in disease-focused genomics research than its predecessor's bulk-tissue approach.